RAG Architecture Skill

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.80). The SKILL.md "Agentic RAG" section defines a tool "search_web" described as "Search the web for current info" and an agent whose system prompt tells it to "Decide which sources to search," indicating the agent will fetch and interpret arbitrary public web content (untrusted third-party pages) and use that content to drive retrieval and action decisions.